Adaptive network for automated first break picking of seismic refraction events and method of operating the same

• US 5,181,171 A
• Filed: 09/20/1990
• Issued: 01/19/1993
• Est. Priority Date: 09/20/1990
• Status: Expired due to Fees

• Current Assignee: Atlantic Richfield Company (BP PLC)
• Original Assignee: Atlantic Richfield Company (BP PLC)
• Inventors: McCormack, Michael D., Rock, Alan D.
• Primary Examiner(s): MacDonald, Allen R.
• Application Number:
  US07/585,967
• Time in Patent Office: 852 Days
• Field of Search: 364/513, 364/421, 364/420, 367/9, 367/14, 367/15, 367/21, 367/24, 367/25, 367/34, 367/36, 367/37, 367/38, 367/54, 367/68, 395/22, 395/929
• US Class Current: 702/14
• CPC Class Codes: G01V 1/288 Event detection in seismic signals, e.g. microseismics G01V1/36 takes precedence
• CPC Class Codes: G06N 3/082 modifying the architecture, e.g. adding, deleting or silencing nodes or connections
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: Y10S 706/929 Geological, e.g. seismology

Abstract

An adaptive, or neural, network and a method of operating the same is disclosed which is particularly adapted for performing first break analysis for seismic shot records. The adaptive network is first trained according to the generalized delta rule. The disclosed training method includes selection of the seismic trace with the highest error, where the backpropagation is performed according to the error of this worst trace. The learning and momentum factors in the generalized delta rule are adjusted according to the value of the worst error, so that the learning and momentum factors increase as the error decreases. The training method further includes detection of slow convergence regions, and methods for escaping such regions including restoration of previously trimmed dormant links, renormalization of the weighting factor values, and the addition of new layers to the network. The network, after the addition of a new layer, includes links between nodes which skip the hidden layer. The error value used in the backpropagation is reduced from that actually calculated, by adjusting the desired output value, in order to reduce the growth of the weighting factors. After the training of the network, data corresponding to an average of the graphical display of a portion of the shot record, including multiple traces over a period of time, is provided to the network. The time of interest of the data is incremented until such time as the network indicates that the time of interest equals the first break time. The analysis may be repeated for all of the traces in the shot record.

Claims

1. A method for operating an adaptive network to determine a first break in recorded seismic data comprising:

operating a seismic source to generate seismic vibrations;

detecting said seismic vibrations at a plurality of surface locations;

recording the detected seismic vibrations as a plurality of traces;

storing data representing said plurality of traces in a memory;

retrieving a selected portion of said stored data representing a first trace corresponding to vibrations detected at a first surface location, and in which a first break occurs, and a plurality of neighboring traces corresponding to vibrations detected at a plurality of surface locations near said first surface location and in substantially the same direction from said seismic source as said first surface location, over a period of time including an analysis time;

presenting said retrieved selected portion of said stored data to input nodes of said adaptive network;

operating said adaptive network to produce an output indicating whether or not the analysis time corresponds to the time at which said first break occurs in said first trace;

responsive to the output of said adaptive network indicating that the analysis time does not correspond to the time at which said first break occurs in said first trace, adjusting said analysis time; and

repeating said retrieving, presenting and operating steps for a portion of said stored data representing said first trace and said plurality of neighboring traces over a period of time including the adjusted analysis time.

2. The method of claim 1, wherein the analysis time in said retrieving step corresponds to a point in time prior to the time at which said first break occurs;

and wherein said adjusting step increments said analysis time to a later point in time along said first trace and neighboring traces.

3. The method of claim 2, wherein the portion of said stored data retrieved in said retrieving step is centered about said analysis time.

4. The method of claim 1, wherein the portion of said stored data retrieved in said retrieving step is centered about said first trace.

5. The method of claim 1, further comprising, responsive to the output of said adaptive network indicating that the analysis time corresponds to the time at which said first break occurs in said first trace;

retrieving a portion of said stored data representing a second trace corresponding to vibrations detected at a second surface location, and in which a first break occurs, and a second plurality of neighboring traces corresponding to vibrations detected at a plurality of surface locations near said second surface location and in substantially the same direction from said seismic source as said second surface location, over a period of time including an analysis time;

presenting said retrieved portion of said stored data to input nodes of said adaptive network, said adaptive network presenting an output indicating that the analysis time corresponds to the time at which said first break occurs in said second trace;

responsive to the output of said adaptive network indicating that the analysis time does not correspond to the time at which said first break occurs in said second trace, adjusting said analysis time; and

repeating said retrieving and presenting steps for a portion of said stored data representing said second trace and said plurality of neighboring traces over a period of time including the adjusted analysis time.

6. The method of claim 5, wherein said first trace is one of said second plurality of neighboring traces.

7. The method of claim 1, wherein said storing step comprises:

arranging said data representing said plurality of traces as a bit-mapped display; and

storing said bit-mapped display arrangement in said memory.

8. The method of claim 7, wherein said data representing said plurality of traces comprises sampled amplitudes of said traces;

and further comprising;

assigning a digital value to each of said sampled amplitudes in said data representing said plurality of traces, said digital values for a trace corresponding to a graphical representation of the trace;

and wherein said presenting step presents each bit of said digital values to an input node of said adaptive network.

9. The method of claim 8, wherein said retrieving step retrieves the data bits associated with the portion of said stored data;

and wherein said presenting step presents the retrieved data bits to the input nodes of the adaptive network.

10. The method of claim 1, further comprising:

prior to said operating step, setting weighting factors in said adaptive network according to a backpropagation method.